Hacker News Reader

Comment by edunteman | Hacker News Reader

Hey HN!

First-time poster here, excited to show you what I've been working on. I'm a recent mechanical engineering grad and self-taught developer. I thought it was a pain in the ass (can we swear on HN?) to walk to a campus computer lab whenever I needed to do 3D CAD or run intensive simulations or code scripts. My less-than-powerful laptop became preventative in my work.

I built a simple-to-use remote desktop system, with popular engineering apps pre-installed, hosted in the cloud. Now, I'm able to run CAD (Autodesk Fusion 360) and development tools (Android Studio) from a $200 device. I think it's silly that the up-front cost of a workstation can prevent software and mechanical engineers from doing their work.

I threw together a free version for your feedback, and I hope to hear your opinions! I'm interviewing for YC W20, so your honest thoughts (and potential usership) would be fantastic.

Cheers! Erik

View on HN

Are you posting user credentials to a free form tool that then emails them to you in cleartext?

https://imgur.com/a/dzF5Kq9

https://imgur.com/a/p3umOwS

This is admittedly correct. The tool sends (edit: sent) to an API that processes your credentials into the user database.

I had not expected submissions to surge as much as they have, and had not built proper encryption into the user signup flow.

I recognize the issues here and am glad you called them out. To respond: 1) I've changed the web form to gather emails as a waitlist until I properly handle new user submissions. 2) I'll develop acceptable user account creation. 3) I'll give current users the ability or change their credentials under the proper system.

My truest apologies for this oversight, and thank you.

What you've done here is absolutely fucking insane and I will never ever trust you and your project and I hope noone else does either. I'm sorry but this is utterly inacceptable. The fact that you need this to be called out is horrifying and your apology is insufficient.

What tech is the remote desktop based on? VNC?

Quite curious about how you keep the latency low enough that it to type in an editor.

I actually have not engineered the streaming portion, so latency is the same as current VNC and RDP protocols.

To get the product together as fast as possible (first line of code written three months ago), I actually did not reengineer the remote desktop protocol. In fact, the current product simply gives you the IP and password info to plug into your Windows Remote Desktop client app.

Future plans are to integrate RDP into the Booste app, to make the experience seem more native.

After that, I'll be implementing the modern H.265 protocol used by videogame streaming companies such as Vectordash and Nvidia GeForce Now.

> I'll be implementing the modern H.265 protocol used by videogame streaming companies

Are you sure that's the right solution?

Remote desktop is able to render sharp fonts because its sending a lot of desktop contextual information over the wire, as opposed to just doing a dump "capture the screen, encode it, and send" like the videogame streaming systems do.

In a videogame, you can get away with small fringes on things (compression artifacts) but it would be quite jarring in your text editor.

I've actually ran into this with the NX protocol (NoMachine) before, and it's one of the main reasons I went away from it.

I agree with this, and it's exactly the reason why RDP is a vastly better experience than VNC.

Using VNC on a LAN isn't great, but over WAN it's a truely miserable experience.

Cool insight here that I was unaware of. I'm still assessing solutions.

It really all depends on which use cases I end up focusing on. The initial idea was for 3D design (IE Solidworks), in which case the H.265 protocol would handle the more graphical, non text-based, visuals. This HN launch was to test out the developer markets, and it seems to resonate enough where I'd consider a way to transmit and render text as text rather than a screen capture.

h.265 will require hardware compression on the encode side (GPU) so you'll have to factor that into the cost.

For sure! Glad to have the tech insight.

I already tether a GPU to the encode side for the 3D design apps, but I'll certainly need to rebuild for scaleability in the future.

Apparently Citrix according to another comment.

I thought it might have been https://www.spice-space.org/ which is GPL2.

Amazon workspaces seems to use RDP

Workspaces uses PCoIP AFAIK

I wish mobile OSs had an implementation of this or similar.

I encourage users to use their Windows RDP client for this version.

Windows Remote Desktop has been usable or at least not completely unacceptable in the past. VNC though I've never seen it be acceptable for CAD so I'm curious too.

Windows Remote Desktop is the current tech in Booste, and is useable but not necessarily pleasant, I admit.

Engineering priority #2, following multi-device-support, is to tackle the streaming challenge.

I'm trying my best, as a solo founder, to not build out features until they're requested, so consider "Faster Streaming" to be requested!

You mentioned somewhere that you are using user-dedicated AWS VMs, but I don't see any mention of what size of instance you are getting? Does it vary depending on the Intermediate/Unlimited plan?

Free plan has the same instance sizing as the other two plans.

If you're technically interested, I serve up a 2xl sized instance, tethered to a medium GPU, regardless of the app used. It's costly to me, so in the future, I'll be finding ways to dynamically size vCPU and vGPU based on workload.

> 2xl

Wow, those are pretty expensive; I'd assumed it would be a medium or large instance!

Hmm, not sure which series of 2xlarge instance you mean, but even with the cheapest that supports Elastic Graphics, doesn't that mean that a user on the Unlimited plan could cost you over $300/month, while you charge them $60/month?

Maybe I'm calculating something wrongly here?

Also, on the Intermediate plan it says "App Sessions Capped At 6 Hours" - what exactly does that mean though? Does it mean I can use it at most 6 hours in a day, or does it mean I get cut off every 6 hours but can immediately reconnect if I want to?

Pricing: my calculations are currently at $30/month per user, at 8hr/day usage with this instance sizing. I'll double-check the sources I used to get pricing.

Medium instances would be appropriate for IDE use cases. I upped to 2xl to handle 3D design tools, but with the recent interest out of the developer community, I can readjust myself downward from 2xl to medium or large.

You get cut off every 6 hours (instance spins down automatically based on your plan). Immediately, you can spin it back up and reconnect for another 6 hrs. The same applies to the free tier, at a 1 hr cap.

Reply on news.ycombinator.com